DEFAULT IN A GOVERNMENT-SPONSORED AGRICULTURAL LOAN PROGRAMME IN SOUTH-EASTERN NIGERIA

Transcription

1 DEFAULT IN A GOVERNMENT-SPONSORED AGRICULTURAL LOAN PROGRAMME IN SOUTH-EASTERN NIGERIA By Jacinta Nwachukwu* Salford Business School University of Salford, UK address: j.c.bottomley@salford.ac.uk Telephone: +44 (0) *This work was financed under a grant from the British Academy. I also wish to express my gratitude to Professor C.C. Asiabaka, Vice Chancellor of the Federal University of Technology in Owerri (FUTO) and to Mr A. Nwaiwu, the Director of the Imo State Supervised Agricultural Credit Loans Board (ISACLB) for their considerable assistance in carrying out the research. I have received judicious advice from Mr E.D. Iwuanyanwu, a loan office from ISACLB. I would also like to thank those who carried out the field work. 1

2 DEFAULT IN A GOVERNMENT-SPONSORED AGRICULATURAL LOAN PROGRAMME IN NIGERIA ABSTRACT Purpose: The Imo State Supervised Agricultural Credit Loans Board (ISACLB) has outright default rates of more than 50 percent. Thus, the purpose of this study is to identify the major characteristics of the Board s beneficiaries who completely failed to honour their repayment commitment as opposed to those who partially repaid. Design/Methodology/Approach: Data on thirty-six potential causes of delinquency were collected through questionnaires distributed to 182 defaulters across ISACLB s three regional zones from 1987 to 1997; ISACLB s only completed loan cycle. Descriptive statistics were obtained using the odds ratio technique. Thereafter, a binary logistic regression estimated the marginal effect on the outright default probabilities of each factor. Findings: ISACLB s large overdue problem was strongly linked to four key factors: (i) age of borrowers, (ii) frequency of visits by loan officers-cum-extension agents, (iii) amount of savings deposits with informal clubs and (iv) total annual savings. Research limitations/ Implications: The primary drawback is the small size of the sample study, as well as the failure to correctly classify our partial defaulters in terms of the stage in the loan cycle at which they actually ceased to repay. Practical implications: In general, initiatives to attract young entrepreneurs, as well as to incorporate a FINCA-type savings scheme into the design of ISACLB s future lending programme should help to resolve its overdue dilemma. Social implications: Older traditional farmers are the principal defaulters. A targeted monitoring, training and information provision appears to be required. Originality/value: There are very few econometric studies dealing specifically with the characteristics of outright and partial microcredit defaulters. Key words: Microfinance institutions, Agricultural finance, Agribusiness, Africa JEL Classification: G21, Q14, Q13, N97 2

3 DEFAULT IN A GOVERNMENT-SPONSORED AGRICULATURAL LOAN PROGRAMME IN NIGERIA INTRODUCTION Microfinance has been heralded as a sustainable way of resolving the problem of lending to poor households with any consequent reduction in poverty and gender inequality. This is because, unlike traditional banks, microfinance institutions (MFIs) have developed innovative mechanisms of lowering their transaction costs and of overcoming asymmetric information problems arising from adverse selection, moral hazard and contract enforcement (Stiglitz, 1990; Besley and Coate, 1995; Ghatak, 1999, Armendáriz and Morduch, 2010). But microfinance programmes are not the first attempt by low-income countries to develop their informal credit markets, promote micro- and small-enterprise development and foster social change. Indeed, in the four decades following World War II, large statesponsored development banks were set up to allocate subsidized credits to socially deprived groups and sectors favoured by the government such as agriculture. The belief was that the extension of cheap loans to farmers would motivate them to adopt new crop varieties and technologies. The subsequent increase in agricultural output should, it was argued, help raise borrower income, employment and welfare and allow for the repayment of the loans. This study aims to contribute to the empirical literature on the credit repayment behaviour of beneficiaries of government-financed agricultural loan schemes. More specifically, it examines the robustness of factors frequently suggested in the extant literature in predicting the probability of an outright default among the borrowers from the Imo State Supervised Agricultural Credit Loans Board (ISACLB). The decision to concentrate on this particular Board s unpaid loans was motivated by the fact that this issue may be more widespread among farmers who borrow from government-sponsored programmes than it is among those who use other sources, including the institutions that reported to the 3

4 Microfinance Information exchange, Inc (MIX). Therefore, the results of our analysis will highlight the critical issues that must be considered by state-sponsored MFIs when screening, disbursing and collecting loans in order to reduce the proportion of their borrowers in complete arrears. To the best of our knowledge, no other study in the microfinance literature has attempted to econometrically analyse the characteristics of beneficiaries that are in arrears with their loans in the credit market of developing countries. This paper is organised under four sections. Section 1 provides an overview of the key institutional structures of ISACLB. Section 2 describes the research method employed in drawing up our sample. Section 3 presents the results of our econometric analysis of the factors that affected outright versus partial default odds among the Board s borrowers from 1987 to 1997, the latest completed loans cycle. The choice of our time period is dictated by the years when ISACLB obtained substantial funds from its sponsors for distribution to agricultural enterprises. However, the Board has recently received a further one billion Naira from the Central Bank of Nigeria. Such has enabled it to recommenced granting new loans of a significant amount upon which this report might bear. To this end, we recommend policies for tackling ISACLB s repayment problems in the final section. 1. INSTITUTIONAL BACKGROUND This Imo State Supervised Agricultural Credit Loans Board was set up by Edict No.1 of the State in But the Board did not start granting loans in reasonable amounts to farmers until 1987 following an amendment in Edict No.10 of that year. This statute empowered the Board to provide subsidized microcredit to individual farmers and cooperative societies as well as operators of agro-allied industries and food processing plants to enable them to establish, reactivate and expand their enterprises. By its own calculations, the Board has been the primary formal micro-loan provider in a State with around 4.8 million inhabitants. 4

5 The operation of the ISACLB is organised into three zones. They are: (i) Owerri with nine local government areas, (ii) Orlu comprising twelve local government areas and (iii) Okigwe containing six local government areas. This makes a total of twenty-seven local government areas in Imo State. The Board s process of screening, granting, disbursing and monitoring of loans across its regional zones was sub-divided into four stages. First, at the zonal level, application forms were obtained at a fee of Naira 2000 in 2011 prices (equivalent to US$13). Applicants were then assigned to a specific loan supervisor who assisted in the completion of their forms. They were pre-screened and interviewed by their appointed loan officers regarding their business plans, household accounts, primary occupation, business registration, insurance and suitability of guarantors in order to establish whether they qualified for a loan under the criteria employed by the Board. Second, loan applications were then subjected to a further review by a technical committee which comprised zonal loan officers and field staff with specialist knowledge of the proposed enterprise and its location. A visit to the applicant s business and household was arranged in order to authenticate the information taken at the preliminary screening stage. The information collected by the zonal project technical committee was organised into a formal loan proposal and presented to the Board s executive body at the State headquarters. Third, if the Board s executive approved a loan application, the list of names of successful candidates was posted in the relevant zones. The disbursement of funds was in the form of a cheque to be cashed at participating commercial banks. This began as soon as the loan agreement was co-signed by two immediate family members of the intended beneficiary as well as by two long-serving senior civil servants who must have been from pay grade 10 upwards. This was the only form of collateral which prospective borrowers were required to provide. In a typical year, the Board disbursed around eighty percent of its loans to individual 5

6 farmers engaged in cassava and rice production, ten percent for fish and poultry farms, seven percent to cooperative societies and the remainder to agricultural-related enterprises such as handicrafts, trading and transport. A monthly repayment schedule started after six months of a moratorium period and was completed in twenty four to thirty six months. A fixed interest rate of ten percent was charged even though this was very unlikely to cover the unit costs associated with the basic interest rate, loan loss provisions and operating expenses. These costs normally total some 25 percent throughout the developing world (MIX, 2011). Fourth, after the disbursement of loans, credit supervisors who also act as extension service agents, made frequent visits to their assigned borrowers businesses and houses to ensure that funds were effectively employed for the specific purposes for which the loans were granted. Borrowers were also reminded of their next repayment date and the amount due. Loan repayments were made either directly to ISACLB s Accounts Department at the State Headquarters in Owerri or through the visiting loan officers on trust that the money will be passed on to it. Borrowers who defaulted were allegedly pressured by the Board s loan recovery task force, but if arrears continued, legal actions were supposedly taken against the borrowers and their guarantors to recover the amounts owed. However, due to the high cost, bureaucracy and time involved in court proceedings, ISACLB only prosecuted a very small number of defaulters, despite the fact that it had a huge outright failure to repay anything at all. This amounted to almost half of its outstanding loans from 1987 to EMPIRICAL RESEARCH METHOD The purpose of this section is to discuss our (i) data sampling, (ii) the empirical model and (iii) control variables and their expected relationship with the default rate. 6

7 2.1: Data Sampling The information used in this study is derived from the repayment record of borrowers with outstanding arrears on loans provided by ISACLB from 1987 to To obtain a fair representation of the characteristics of defaulters, a two stage sampling procedure was employed. The first step involved a deliberate selection of the three zones in which the Board operates. The second comprised a random selection of 100 beneficiaries from each of the three zones. The total initial sample therefore contained 300 overdue loans made to individual farmers and cooperative societies. The data itself on the factors influencing loan repayment was collected with the help of a questionnaire. The respondents were asked a range of questions aimed at understanding the major reasons why they failed to repay their loans at the specified due dates. The data that was then obtained was checked for errors and omissions to ensure its authenticity and appropriateness. A complete dataset on 36 variables was extracted for a random sample of 182 beneficiaries from each of the Board s three zones. These comprised attributes relating to borrowers personal and household characteristics, the nature of their business, loans and lender characteristics, membership of group savings and cooperative societies, welfare and the adequacy of infrastructure (see Appendix Table 1). As shown there and in Figure 1, our overall sample of 182 respondents were broadly classified as partial or outright defaulters depending on whether they made some or no contribution to their loan repayments over the ten years of the study. Partial defaulters constituted 51 percent of our overall sample (i.e., 93 out of 182 respondents), meaning that the remaining 49 percent (89 out of 182 respondents) completely failed to honour their repayment commitments. This contrasts sharply with Oke et al s (2007) claim that overall default rates among farmers in Nigeria was much lower. 7

8 In addition, a sub-division of our respondents according to the zones in which they resided showed that a typical borrower from the Orlu district was more likely to totally default than partly repay his/her loan to ISACLB. This might have been because the Board s procedures for screening, monitoring and enforcement were inconsistently applied across the different regional zones. Also, it was more likely that loan officers found it easier to visit the residence and businesses of the more affluent borrowers in the urban area of Owerri where it s headquarter is located. 2.2: Model Specification A number of empirical papers have investigated the microcredit repayment rate of farmers in Sub-Saharan African economies in general and Nigeria in particular (Temu 1999, Chirwa, 1997; Njoku 1997; Olomola 2001; Oke et al, 2007; Udoh, 2008; Papias and Ganesan, 2009, Afolabi, 2010, Ojiako and Ogbukwa, 2012). Generally, these studies showed that the default risk of agricultural borrowers may be linked to causes associated with their household, 8

9 business, loan and lender characteristics, as well as their access to adequate infrastructure. We deal with these here by estimating a Multivariate Logit model specified as follows: where is a binary variable with values reflecting the non-repayment status of our survey respondents 1. It takes a value of 1 if an individual completely defaulted on his/her loan repayment (i.e., outright defaulters) and 0 if an individual repaid with arrears (i.e., partial defaulters). The term is a vector of dichotomous variables for each of the zones in which our respondents resided. The dummy for the Urban Owerri zone is the category omitted from the regression. Thus, the differential coefficients on the dummies for Orlu and Okigwe zones account for the possibility that the risk of total default is higher among borrowers in rural areas than in Owerri town. The null hypothesis to be tested is that there are no regional variations. As a result, the expected value of the differential intercept coefficient is equal to zero for both the ORLU and OKIGWE variables. The symbol is a vector of control variables and is a stationary error term Control Variables and Expected Relationships with the Probability of Outright Default Items 2.1 to 7.3 in Appendix Table 1 and in Figure 2 present the summary statistics for the attributes which are considered as control variables in our Multivariate Logit regressions. To be able to initially describe the effect on the outright default probability of a unit change in each of the selected variables in isolation, we calculated the odds ratio in Column 5. For 1 The list and definition of explanatory variables is provided in Appendix Table 3 9

10 brevity, our arguments here are restricted to those variables whose estimated odds ratio differ by at least 50 percent from the default neutral value of one (see Column 6) : The borrowers personal and household characteristics included in our dataset are listed in items 2.1 to 2.7 of Appendix Table 1. Column 6 shows that a borrower age of fifty or over is the most important cause of ISACLB s high default rate, judging by the magnitude of its estimated odds ratio. Older borrowers, it seems, are more likely to be in the total nonpayment than in the partial repayment category. One possible reason for this is that with age comes wisdom. The older borrowers have learned that there are unlikely to be any sanctions for those who do not repay their loans as ISACLB has successfully prosecuted only a very few defaulters since its establishment in Then too, as noted by Olomola (2001), older borrowers are more likely to live in rural areas, engage in traditional farming and to sell their produce in nearby local markets than are younger ones. Moreover, older borrowers are less likely to spread their investments across a variety of activities and so lose out on the opportunity to diversify income covariance risk (Oke et al 2007). We may therefore infer that, to a large extent, ISACLB s high outright default rate is correlative with the fact that the aged fifty and over who make up around fourfifths of its borrowers tend to generate lower net returns with an attendant reduction in their capacity to repay vis-á-vis the young. Consequently, the coefficient on the AGE variable in our estimated model is expected to have a positive sign. 2 Column 3 of Appendix Table 3 provides information on the hypothesised impact of each of our 36 variables on the likelihood of an outright default by a typical borrower in our survey. 10

11 2.3.2: Business characteristics are outlined in items 3.1 to 3.7 in Appendix Table 1. One feature stands out as a major influence on the possibility that our respondents will completely fail to honour their repayment obligations under this category. Contrary to the ISACLB s remit, borrowers who claimed that farming was their only occupation constituted a mere 38 percent of our overall sample. Figure 2 shows that the odds ratio for these wholly-farming respondents was 1.64, indicating a 6.4 percent increase in the risk of outright default for an additional ten percentage point rise in the proportion of completely-farming borrowers. Given that the majority of our borrowers were involved in other enterprises in addition to farming, this result signifies that incomes from non-farming businesses and occupations such as food processing, transportation, teaching, the civil service and the like were used for loan repayments during periods when arable cropping and other farming activities might have experienced financial distress and generally low incomes. The coefficient on FARMG is therefore predicted to be significantly positive : The loans and lender characteristics which are considered in this paper are shown in items 4.1 to 4.5 of Appendix Table 1. Column 6 indicates that visits by loan officers-cumextension agents have notably affected the outright default probabilities of our respondents. A ten percentage rise in the share of our borrowers who benefited from these visits appears to have reduced the odds of total non-payment by 5.8 percent. The possibility of an outright delinquency was diminished by a further 3.7 percent as a consequence of a ten percentage rise in the fraction of our borrowers who were visited more than six times a year. This finding is in line with the empirical evidence provided by Olomola (2001) for Nigeria and Papias and Ganesan (2009) for Rwanda. As argued by these authors, more frequent visits by extension agents and/or loan supervisors helped borrowers raise the yield on their economic enterprises 11

12 and to understand the virtue of maintaining a good credit record. Accordingly, the coefficients on VISITS and NUMVISITS are envisaged to be significantly negative. Figure 2: The Odds Ratio (for belonging to the Outright Default Category) Source: Author s calculations; N = Nigerian Naira, K = Thousands Size of loans disbursed by the ISACLB is another key factor influencing its outright default risk. On average, only 18 percent of loans offered by the Board exceeded 20,000 Naira in constant average 1987 and 1997 prices. The risk of total default was estimated to fall by 5.2 percent for a ten percentage rise in the ratio of our respondents who benefited from this larger loan amount. As noted by Arene (1992), larger loan sizes enable borrowers to increase their business investments and generate higher net returns. Besides, beneficiaries of larger loan sizes are more likely to try new and innovative ventures which are more 12

13 profitable. Further, an allocation of the loans between traditional and modern activities could act as an insurance against the risk of co-variation between incomes from similar productive enterprises. It would seem that improvement in the repayment ability of these larger loan borrowers more than outweighed the additional burden imposed by any greater debt obligation. As a result, it is proposed that the coefficient on LONSIZE will be negative : Membership in rotation or group savings and cooperative societies is used here to account for the differences in the outright default risk of our respondents who belonged to savings clubs and credit cooperatives by comparison with non-members. Participants in an informal rotational or group savings club were reported to have had an odds ratio of This indicated that a 5 percent decrease in the outright default probability was related to a ten percentage rise in the share of informal savings club members. We may therefore infer that savings clubs gave members access to communal money which can be used to service loans when they fall on hard times (Armendáriz and Morduch, 2010). Hence, the variable MBSAVG is conjectured to have a negative coefficient. Deposits with savings clubs were another important determinant of ISACLB s outright default possibility. A ten percentage point addition in the share of our savings club members who made regular deposits lowered the likelihood of total default by 5.8 percent, rising to 8.2 percent for those respondents who deposited 20,000 Naira per annum and above in constant 1987 to 1997 average prices. A savings club, it would seem, offers participants the opportunity to accumulate loan repayment money outside their households and away from a demand from kin and from theft (Gugerty, 2007; Anderson et al, 2009). Also, the public nature of the meetings involved served as a commitment device to foster discipline and compel members to make at least initial part payments to the Board. However, proponents of 13

14 the precautionary theory argue that a relatively high savings rate may reflect the extent of income variability faced by our borrowers. It may well be that prudent farmers acted with foresight and saved when income was high and withdrew their savings to preserve desired consumption levels and to honour loan repayment obligations during downturns. We tested the statistical significance of the hypothesis that the SVDEPST and SVDAMT variables in our model were negatively correlated with the probability of an absolute non-payment The human welfare indicators which are considered here are listed in items 6.1 to 6.6 of Appendix Table 1. Total savings in a typical year have the strongest effect on default risk with an odds ratio of This means that the chances of an outright default were lowered by 7.2 percent for a ten percentage increase in the ratio of our respondents who saved, on average, 20,000 Naira and above a year. Thus, it is hypothesised that the variable SAVGAMT is negatively correlated with the risk of outright non-payment. Such is indicative of the direct relationship between loan repayment and the average income of borrowers (see also Njoku, 1997). The fact that these relatively well-off borrowers failed to repay all their outstanding loans by the end of our ten year cycle suggests that they were opportunistic defaulters who considered government-sponsored loans as a gift. Their decision to default on ISACLB s loans was likely to have been more tactical than an inability to repay. A further key reason for the repayment performance of our borrowers was the availability of collateral assets. Surprisingly, two-thirds of the respondents to our survey claimed to have collateral which could be offered in support for their loans. The outright default among this group of borrowers was 66 percent lower than it otherwise would have been without it, indicating a possibly negative coefficient on COLATRAL in our model. 14

15 What is more, the Board s outright default risk was estimated to have declined by 5.1 percent for every ten percentage point increase in the number of borrowers with land as a guarantee compared to those with other forms of security such as jewellery, livestock, growing crops and the like. We therefore surmise that FMCOLTRL was negatively correlated with an outright default risk. Land-owning borrowers, it seems, are the most likely to fear appropriation. The question therefore arises as to why the ISACLB has been unable to seize the assets pledged as collateral by its defaulting clients. This is doubtless due to inertia and political opprobrium, as well as the cost and time involved in recovery through the courts. Indeed, as reported by Dercon (1999), the nature and value of the assets used by African households for savings and collateral is positively related to incomes. We may therefore deduce that the outright land owners in our survey are more likely to belong to the political elite and/or to be relatively well-to-do and any attempt by the ISACLB to appropriate their properties would then meet with effective resistance. Then too, the Board may have faced added difficulties in rural areas where most land is tribally owned and apportioned. Besides, under its charter, the ISACLB is responsible for the well-being of farmers in the State. The decision not to seize collateral assets may be regarded as part of discharging this responsibility to its borrowers. The majority of our respondents claimed to spend a large percentage of their family income on food, reflecting the Board s anti-poverty mission. It is reported that a further ten percentage point rise in the proportion of borrowers who claimed that food absorbed the largest share of their household income was correlated with a 6.6 percent increase in the probability of being classified as an outright defaulter rather than a partial non-payer. This is suggestive of a likely statistically significant positive coefficient for the FOOD variable. 15

16 Infrastructure-related causes of the defaulting status of borrowers are provided in items 7.1 to 7.3 of Appendix Table 1. As expected, the better these were in terms of storage facilities, electricity supply, transport and communication provision, the less likely was a borrower to default totally on his/her loan repayment. The most marked characteristic in this category is the adequacy of storage facilities with an odds ratio of This implies a 5.1 percent decline in the outright default rate for a ten percentage point increase in the proportion of borrowers who had access to storage amenities which were compatible with their type of business enterprise. Such may have been a reflection of an improvement in the ability of our borrowers to generate higher and more stable cashflows by selling their produce out of season at higher prices as well as by reducing waste. Thus, it is expected that the coefficient on STORG will be significantly negative. The analysis in this section presents a descriptive account of the major factors responsible for the large level of unpaid loans suffered by the ISACLB from 1987 to But the simple, stylized nature of the underlying non-parametric statistics has not allowed for the possible interactions that exist between these explanatory variables. An econometric study is now required to establish the link between outright default probability and the several characteristics enumerated in items 1 to 7 in Appendix Table 1. The technique chosen to do this is the Binary Logit Model which is traditionally used for analysing conditional data. The results from this logistic regression form the basis of our discussion in the next section. 3. Empirical Results and Discussion Appendix Table 2 presents the results of our logistic regression in terms of the marginal effect of each of our explanatory variables on outright default probabilities. Three alternative 16

17 regressions were run in order to check the robustness of our estimated parameter coefficients and, by inference, the odds ratios which formed the basis of our descriptive analysis in the previous section. Our summary of the key findings is as follows: 3.1: The Marginal Effect of Regional Characteristics The first logistic regression in Column 1 excluded all of the independent control variables. The results showed that a ten percentage point increase in the proportion of borrowers from urban Owerri would reduce ISACLB s risk of total non-payment by 1.83 percent. With a similar increase in the share of borrowers from the rural Orlu and Okigwe branches, the threat of an outright default was raised by 2.95 percent and 0.6 percent respectively above the rate reported for their counterparts in Owerri, although the implied difference for the Okigwe zone was statistically equal to zero at the conventional 5 percent level. An assessment of the overall fit of this simple model using the Schwarz Bayesian Information Criterion the McFadden Pseudo R-squared and the Chi-squared test statistic suggests a low explanatory power for our group of regional dummy variables by themselves. Additionally, the model s success ratio of correctly classified cases of outright and partial defaulters is percent assuming a 0.5 cut-off point. This is significantly below the correct classification ratio of circa 73 percent reported for the alternative regression models in Columns 2 and : The Marginal Effect of Borrowers Personal and Household Characteristics Regression 2 in Column 2 shows the marginal impact of the three regional dummies plus the thirty-six control variables as specified in equation 1. This alternative model explains percent of the variation in the probability of total non-payment among our borrowers. As a 17

18 result, the percentage of our respondents accurately classified as belonging in the outright or partial defaulting category rose to percent compared to the percent reported in regression 1. Column 3 indicates that the results of a parsimonious regression which excluded all of the statistically insignificant variables were broadly consistent with those of our extended model. Consequently, much of our argument in the rest of this section is derived from the estimated coefficients and their asymptotic t-statistics resulting from the expanded Multivariate Logit model in Column 2. With regard to the category comprising borrower-related causes, four of the variables, namely GENDER, MARITAL, HEALTH and EDUCBI, have the expected negative coefficients while the remaining three, AGE, NUCLEAR and EXTEND, have positive coefficients. However, the t-statistics associated with these estimated coefficients indicate that the effects on total default possibility of changes in all of these characteristics are insignificantly different from zero, with the exception of AGE. A ten percentage point increase in the ratio of the older to younger borrowers would raise ISACLB s outright default rate by 4.18 percent. The implied adverse effect of a change in this variable on the chances of outright default is markedly lower than the 11.5 percent suggested by our non-parametric odds ratio in the previous section. This is presumably a reflection of the interrelatedness between a borrower s age and the other attributes in our expanded regression model. For example, older borrowers are more likely to be married men. They are therefore able to mitigate somewhat the perverse effect of the age factor. An assessment of the diagnostic statistics indicates that this extended regression provides a good fit to our data. 18

19 3.3: The Marginal Effect of Business Characteristics The variables included in this category with negative coefficients are BUSOWN, FARMG and FCPRICES. By contrast, the remaining four factors that have a positive effect on the absolute default rate of borrowers are BUSTYPE, TRAING, REKEPS and DISTNCE. The effects of all these business-related attributes are insignificantly different from zero at the five percent test level. Nonetheless, it is noteworthy that the positive coefficient on DISTNCE is in conformity with our expectation. In explanation, one of ISACLB s loan supervisors commented that borrowers whose businesses were closer to their homes were able to move with relative ease between the two registered addresses and hence to successfully dodge them whenever they came to collect repayment. Further, it was remarked that enterprises that were closer to home, such as bee keeping, poultry, fisheries and snailries, were less profitable than those further away from home, such as rice, palm oil and rubber plantations. 3.4: The Marginal Effect of Loan and Lender Characteristics The five characteristics which are included in this category have the expected negative effect on outright default likelihood, even though the coefficient on NUMVISITS is the only one that is statistically different from zero. A ten percentage point rise in the proportion of borrowers who were visited by extension agents more than six times a year was related to a decline of 2.93 percent in the possibility of complete non-payment. We may therefore infer that the favourable impact of the frequency of visits on the risk of outright default is indicative of the success of the ISACLB s staff in monitoring and advising its borrowers. 19

20 3.5: The Marginal Effect of Membership in Rotation or Group Savings and Cooperative Societies Four out of the eight variables which we included in this category had the predicted negative coefficients demonstrating their favourable impact on the probability of total non-payment. They were: MBSAVG, SVDAMT, MBCORP and ATMEETG. However, the only factor which had a significant influence on default odds was SVDAMT. A ten percentage point rise in the share of savings club members with savings of 20,000 Naira and above in average 1987 to 1997 constant prices reduced the Board s outright default rate by 4.70 percent. The positive coefficient on our four other characteristics of savings and cooperative society participation, SVDEPST, CORPDEPST, CORPDAMT and LEADRQ, indicated that changes in these variables would raise the odds of total default for our typical borrower. In particular, there was evidence that a ten percentage point increase in the share of savings club or cooperative society members who claimed to have good/excellent leaders significantly raised their outright default probability by 3.37 percent. Although this result is contrary to experience elsewhere (Cassar et al, 2007), it is possible that borrowers were informed by particularly persuasive leaders that they were unlikely to be punished by ISACLB on default and were judged accordingly by members. Then too, good/excellent leaders are more likely to enforce and penalize members who break the rules and regulations of the savings and cooperative unions. High penalty charges will reduce the amount of money available for loan repayment and might also discourage regular attendance at group meetings (Oke et al, 2007). Our results indicate that borrowers who attended group meetings less than six times in a year had higher outright default odds. In general, our results indicate that membership of an informal savings group was more effective in solving ISACLB s huge loan overdue problems than it was in a cooperative society. Such hints that there are further constraints imposed on borrowers who participate in 20

21 a cooperative society vis-á-vis those who choose to belong to an informal savings clubs. For example, cooperative societies perform other functions such as purchasing inputs and marketing output. Such extra activities may reduce the amount of loanable funds available to members who wish to borrow for investment and/or for servicing outstanding loans from outside providers. Then too, the extent of collaboration, trust and social ties among informal savings club participants is stronger than with cooperatives since they are more likely to be friends, relatives and acquaintances. Such informal arrangements are more likely than formal cooperative societies to successfully overcome asymmetric information, develop strong ties between participants and provide incentives to save and repay. 3.6: The Marginal Effect of Human Welfare Characteristics The estimated coefficients on the majority of the human welfare attributes were negative, although their t-statistics indicated that SAVGAMT was the only factor likely to significantly influence the outright defaulting decision of borrowers. This was to be expected as those who had cultivated the discipline to save 20,000 Naira or more in a year were most likely to be those who recognized the importance of honouring commitments to finance providers so as to gain access to cheaper loans in the future, even though they did not persist in this belief. They were also more likely to be better educated and hence able to identify the various savings opportunities offered by the banks and to understand that past ravaging inflation in Nigeria significantly reduced the real value of money repaid to ISACLB. The fact that this group of borrowers eventually defaulted suggests that such knowledge was again eventually outweighed by the absence of sanctions. 21

22 3.7: The Marginal Effect of Infrastructure Characteristics The negative coefficient on STORG is in conformity with the hypothesis that enhancing the quality of storage facilities available to our borrowers should go hand-in-hand with an improvement in their repayment ability. On the other hand, contrary to expectations, having access to a regular supply of electricity and a good road and telecommunications network appears to raise the likelihood of absolute default. But the fact that none of our infrastructurerelated variables are statistically significant suggests that their effectiveness in mitigating the risk of total default is already embodied in our other control variables. For example, good roads, electricity and telecommunication systems will facilitate access to storage and urban markets, as well as up-to-date information on the prices therein. Further, they are likely to correlate with the number of visits by extension agents and loan officials. An increase in the frequency of such visits is significantly related to a reduction in the risk of total default. CONCLUSIONS AND POLICY RECOMMENDATIONS This paper has provided evidence on the factors which influenced the likelihood of outright default among borrowers who received loans from the Imo State Supervised Agricultural Credit Loans Board between 1987 and The results of a Multivariate Logistic regression indicated that the Board s large loan overdue problem might be linked to four factors, judging by the fact that their estimated coefficients were persistently statistically significant at the five percent level in both our extended and parsimonious models. First, age is the most important borrower-related cause of outright default among our respondents. A further ten percentage point increase in the fraction of borrowers aged 50 years and above raised the likelihood of absolute non-payment by 4.18 percent. But data also 22

23 indicated that older borrowers were more likely to have relatively low incomes, to be in poor health, less well educated, to be rural dwellers, full time farmers and to keep no record of their inputs and outputs. These characteristics lay behind the perverse influence of age on total default risk. Nevertheless, an obvious solution to this problem is to increase the percentage of total loans extended to the young who comprised a mere 20 percent of ISACLB s borrowers. Initiatives to attract the young may include advertising on television, newspapers and social network websites such as Facebook, as well as visits to institutions of secondary and tertiary education. Another way out of the Board s overdue problem would be to raise the share of loans offered to those older educated, male, married borrowers with a higher number of younger extended versus nuclear family members who reside with them. There is evidence that wives and extended relatives may help raise the overall household income and loan repayment therefrom as they are generally put to productive work. Second, the frequency of visits by extension agents, who also appear to be the loan officers themselves, is the only loan and lender characteristic significantly affecting default odds. A rise of ten percentage points in the share of loan beneficiaries visited more than six times a year reduced the chances of outright default by 2.93 percent. Unfortunately, only 35 percent of the ISACLB s clients were visited this often, presumably because of poor roads and lack of vehicles. This indicates that initiatives which improve ease of access and communication with elderly borrowers in rural areas in particular might significantly lower the Board s high outright default rate. Such actions may include the recruitment of more loan officers-cum-extension agents to work in the Board s rural offices, the provision of four wheel drives and/or less expensive motorcycles with fuel for visiting loan officers, as well as improving road and track networks for officer visits and the marketing of produce. Then too, guidance pamphlets which deal with problems frequently encountered by borrowers engaged in different business enterprises for which the Board s funds are normally deployed should be 23

24 issued to all successful loan applicants at the start of each loan cycle. Regrettably, these solutions will involve additional lending costs and consequent higher interest rates on loans. At the moment ISACLB is not permitted to charge more than 10 percent on its loans, whereas data from the Microfinance Information exchange (MIX) shows that some 25 percent net of inflation is required to cover lending costs. Interest policy needs, therefore, to be realistic and/or the question of the comparative advantage of subsidized microlending addressed. Third, savings club deposits of 20,000 Naira and above in constant 1987 to 1997 average prices were the most important cause of absolute default risk associated with membership of rotational savings or cooperative societies. An additional ten percentage point increase in the proportion of savings club contributors with deposits of 20,000 Naira or more would reduce the prospect of complete non-payment by 4.70 percent. This may be indicative of the favourable impact of trust and social ties among participants of such informal social arrangements. However, only 30 percent of the Board s borrowers had deposits with savings clubs in excess of this threshold. Thus, the incorporation of a corresponding savings scheme into the design of ISACLB s future lending programme could be instrumental in resolving some of its overdue dilemma. An implementation of such savings plan may require that the Board adopts the village banking lending approach first introduced extensively by FINCA International in A FINCA-type village banking scheme normally includes between 10 and 30 people in a group who meet regularly at a designated community centre under the supervision of its loan officer. Each week, members are required to make a fixed instalment payment, comprising repayment on their loans and savings deposits. In the event of default, a similar group s savings would be used to pay back ISACLB. Monitoring and enforcement could be improved by ensuring that all payments are accurately recorded and by producing posters propagating the virtues of camaraderie, trust and teamwork in building a successful business. To increase accountability and motivate 24

25 staff, ISACLB should set up an incentive scheme linking employee bonuses to recognizable performance indicators. Such might reward officers who do extremely well in assessing, extending and collecting savings deposits and weekly loan repayments. Fourth, with regard to human welfare-related causes, the negative coefficient reported for SAVGAMT was the only statistically significant item. For an extra ten percent point rise in the proportion of borrowers with savings of 20,000 Naira and above in constant average 1987 to 1997 prices, the threat of outright default declined by 2.33 percent. Given that only 43 percent of the Board s borrowers had savings above this cut-off point, FINCAtype village banking initiatives to encourage participation in informal savings clubs along the lines described above should help lower outright default rates. Then too, banks in which such savings may be deposited are often seen as forbidding and otherwise difficult to deal with. In this respect, where ISACLB transacts with the commercial banks, its officials may try to advise them on, and guide reforms in the provision of services for smallholder farmers. Then too, the interest rates which banks pay on their deposits net of inflation are commonly very low or even negative. But this is a national problem and ISACLB can only join in lobbying for improvement in this respect. An important limitation of this study is that generalisation for policy reforms for ISACLB and other government-sponsored microlending institutions is based here on a survey of only 182 respondents with outstanding arrears on loans provided by the Board from 1987 to A major area for further research therefore would be to considerably increase this sample size and to update our dataset by questioning borrowers who have benefited from the Board s recent lending in the 2011 to 2012 loan cycle. Another drawback stems from the fact that information on the income generated by the various enterprises to which the Board s loans were put was missing, perhaps because of poor record keeping and the reluctance of respondents to give full and accurate account of their earnings. Then too, our estimated model 25

26 assumes that all partial delinquents dropped out of the loan cycle at the same point. A Multinomial Logit model which reclassifies our partial non-payers in terms of the stage in the loan cycle when they actually defaulted would help the ISACLB target resources more accurately, in particular visits by its loan officer-cum-extension agents. References Afolabi J.A (2010), Analysis of Loan Repayment among Small Scale Farmers in Oyo State, Nigeria Journal of Social Sciences, Vol. 22 (2), pp Anderson, S., Baland,J-M., Moene, K.O (2009), Enforcement in Informal Saving Groups Journal of Development Economics, Vol. 90, (1), pp Arene, C.J (1992), Loan Repayment and Technical Assistance among Smallholder Maize Farmers in Nigeria, African Review of Money, Finance and Banking, Supplement of Savings and Development Journal, Vol. 1, pp Armendáriz de Aghion, B., and Morduch, J (2010), The Economics of Microfinance, Cambridge, MA: MIT Press Besley, T., and Coate, S (1995), Group Lending, Repayment Incentives and Social Collateral, Journal of Development Economics, Vol.46 (1), pp.1-18 Cassar, A., Crowley, L., and Wydick, B. (2007), The Effect of Social Capital on Group Loan Repayment: Evidence from Field Experiments, Economic Journal, Vol.117 (517), pp. F85-F106 Chirwa, E.W. (1997), An Econometric Analysis of the Determinants of Agricultural Credit Repayment in Malawi, African Review of Money, Finance and Banking, Supplement of the Saving and Development Journal, Vol. 1, No. 2, pp Dercon, S (2002), Income Risk, Coping Strategies, and Safety Nets, World Bank Research Observer Vol. 17(2): pp Ghatak, M (1999), Group lending, Local information and Peer Selection, Journal of Development Economics, Vol. 60 (1), pp Ghatak, M., and Guinnane, T (1999), The Economics of Lending with Joint Liability: Theory and Practice, Journal of Development Economics, Vol. 60 (1), pp Gugerty, M.K (2007), You Can t Save Alone: Commitment in Rotating Savings and Credit Associations in Kenya, Economic Development and Cultural Change, Vol. 55, No. 2, pp

27 Njoku, J.E (1997), Determinants of Loan Repayment under the Special Emergency Loan Scheme (SEALS) in Nigeria: a Case Study in Imo state, African Review of Money, Finance and Banking, Saving and Development Journal, Vol. 1, pp Ojiako, I.A., and Ogbukwa, B.C (2012), Economic Analysis of Loan Repayment Capacity of Smallholder Cooperative Farmers in Yewa North Local Government Area of Ogun State, Nigeria, African Journal of Agricultural Research, Vol. 7 (13), pp Available online at Oke, J.T.O., Adeyemo, R., and Agbonlahor, M.U (2007), An Empirical Analysis of Microcredit Repayment in Southwestern Nigeria, Humanity and Social Sciences Journal, vol.2 (1), pp Olomola, A.S. (2001), The Nature and Determinants of Rural Loan Repayment Performance in Nigeria: The case of FADU s Microcredit Programme, NISER monograph series No. 3, NISER, Ibadan. Papias, M.M., and Ganesan, P (2009), Repayment behaviour in credit and savings cooperative societies: Empirical and theoretical evidence from rural Rwanda, International Journal of Social Economics, Vol. 36, No. 5, pp Stiglitz, J.E (1990), Peer Monitoring and Credit Markets, World Bank Economic Review, Vol. 4 (3), pp Temu, A.E. (1999), The Kilimanjaro Cooperative Bank: a Potentially Sustainable Rural Financial Institutional Model for Sub-Saharan Africa, African Review of Money, Finance and Banking, Supplement for the Savings and Development Journal, Vol. 1 No. 2, pp The Mix Market Database (2011), Microfinance Information exchange, Inc (MIX), available on the website: Udoh, E.J (2008), Estimation of Loan Default among Beneficiaries of a State Government Owned Agricultural Loan Scheme, Nigeria, Journal of Central European Agriculture, Vol. 9 (2), pp